Subtyping Chronic Kidney Disease Patients And Adiposity-Obesity Related Metabolomics Analyses: Findings From The Chronic Renal Insufficiency Cohort Study

Zihe Zheng, University of Pennsylvania

Abstract

Chronic kidney disease (CKD) is a heterogenous condition that is often complicated by multiple serious comorbidities that create a large disease burden. Concurrent with the high CKD prevalence is the epidemic of obesity which increases the risks of adverse outcomes among people with kidney dysfunction. However, due in part to patient heterogeneity, the complex relationship between obesity and CKD is not fully understood. We aim to systematically examine phenotypic heterogeneity in patients with CKD and to study CKD mechanisms related to obesity-adiposity by integrating rich clinical characteristics of patients with high-dimensional metabolomics data. 3939 participants in the prospective Chronic Renal Insufficiency Cohort (CRIC) Study with stage 2-4 CKD at baseline were included in this body of research. We conducted two parallel clustering analyses using the machine learning methods of consensus clustering. First, we examined the overall CKD heterogeneity using 72 markers of patients’ demographics, biomarkers, and commonly collected clinical characteristics. Second, we identified the adiposity-obesity-related (AOR) CKD subgroups using 22 markers of patients’ obesity attributes, adiposity parameters, and comorbidity profiles. Third, in a random subset of CRIC participants with metabolomics data, we investigated the metabolic signatures associated with AOR CKD subgroups and tested metabolites as potential mediators of the association between AOR CKD subgroups and various clinical endpoints using Aalen additive hazards models and Cox regression. Among our findings, we identified three distinct CKD subgroups from the overall clinical data, and a different set of three-level AOR CKD subgroups featured with distinct patient profiles of adiposity/obesity and diabetes. Both sets of CKD subgroups were significantly and independently associated with different rates of future clinical outcomes. The metabolomics and mediation analyses revealed numerous metabolites to be mediators of the relationship between AOR CKD subgroups and clinical endpoints. Among them, multiple lipids, nucleoside, and amino acid metabolites were identified as key markers. In summary, our work quantitatively characterized CKD patient heterogeneity, shed light on adiposity-obesity-related disease mechanisms at both phenotypic and molecular levels, and highlighted potential therapeutic targets as well as metabolomics pathways for disease management and treatment. Validation using longitudinal metabolomics data and/or independent cohorts are needed.